Upload
vuongdien
View
214
Download
0
Embed Size (px)
Citation preview
MPRAMunich Personal RePEc Archive
On the Investigation of Factors EffectingInternational Tourist Arrivals toCambodian Market: A Static andDynamic Gravity Approach
Theara Chhorn
Faculty of Economics, Chiang Mai University (CMU), Thailand
24 April 2017
Online at https://mpra.ub.uni-muenchen.de/83813/MPRA Paper No. 83813, posted 12 January 2018 09:18 UTC
On the Investigation of Factors Effecting International Tourist
Arrivals to the Cambodian Market: Static and Dynamic Gravity
Approach
Theara Chhorn a
a Faculty of Economics, Chiang Mai University (CMU), Thailand
Please cite this article as:
Theara C. (2017). On the Investigation of Factors Effecting International Tourist Arrivals to the
Cambodian Market: A Static and Dynamic Gravity Approach. Southeast Asian Journal of Economics
5(2), pp. 115-138
Abstract
The paper investigates factors influencing international tourist arrivals into the Cambodian market
during the period of 1995 to 2015, covered 32 cross-sectional countries by adopting a static and dynamic
gravity approach with respect to fixed effects (FE) and random effects (RE) and the GMM estimator of
Arellano and Bond (1991). Our analysis shows that mostly economic factors such as travel cost, GDP per
capita and population size are the main sources in attracting international tourist arrivals. The country
specific dummy variables are found to be associated with the respects to its significant level. The empirical
results demonstrate that one and two step GMM with robust standard errors produces better outcomes and
improves the estimated accuracy over a static approach.
Keywords: Gravity Model, International Tourist Arrivals, Arellano and Bond (1991), GMM,
Fixed and Random Effect, Cambodia
JEL Classification: C23, C50, Z30, Z32
1. Introduction
Tourism sector so-called the non-smoking industry played an essential function in
boosting and sustaining socio-economic development in the single market trend generating
from in the era of modern globalization. It is furthermore considered as one of the most
potential sector which contributes to both household’s welfare and national income
throughout, economic growth, employment creation, global and regional connection in every
corner of the world. Recently, the total contribution of travel and tourism to GDP was
accounted almost 9.8% of GDP and is expected to rise by 3.7%, equaled to 9.9% of GDP in
20151. More importantly, the flow of tourism demand between the origin country and
destination country is depended on tourism related activities and policies of international
corporation relevant to socio-economic stability, low risk in internal and external violent in
the country such as terrorism attack, to which impulse tourist’s manner and decision.
In the context of Cambodia, tourism sector is reflected as the most driven factors to
social and economic development. For instant, the area of Angkor temple was listed in the
world heritage in 1992, as well as the fresh opening economy into regional and international
system since 1993 up to present time, it was such the main catalyst without catastrophe to
encourage international tourist arrivals. Beside these, the number of tourist arrivals has
increased dramatically and risen year-on-year. Statistically, it was contributed 29.9% to GDP
in 2014 and expected to augment from 8.2% to 30.2% of GDP in 2015. Moreover, it is
forecasted to reach up by 6.5% per year equivalent to (28.0% of GDP)2 by 2025. So far,
Cambodia’s government puts the eyes strongly to develop such growing sector, due to the
facts that it is not only contributed directly and indirectly to socio-economic development and
economic growth, poverty reduction but it is also promoted the country’s appearance and
landmark into the world.
1
WTTC Travel and Tourism Economic Impact 2015, WTTC stands for World Tourism and Travel Council
2 WTTC Travel and Tourism Economic Impact 2015, CAMBODIA
Figure 1: Number of international tourist arrivals by 6 origin countries (1995-2015)
Series ASEAN East Asia Southern Asia Oceania European American
Average 115670 199343 6814 34170 121584 71006
Median 113335 192419 6050 36058 118447 77009
Source: Computer calculation, data extracted from CIEC database manager
From 1995 to 2015, most of the tourism flow is somehow attracted by ASEAN region
and ASEAN partnerships such as China, Japan and Korea followed by European and
American. Vietnamese tourist is ranked the number 1st for tourism industry in Cambodia
since 2010, China is represented as the 2nd largest followed by Korea and Thailand. This
movement was reduced since 2005 due to the global financial crisis in 2008 and 2009 as well
as external violent within the neighboring country, say Thailand relevant to PREAH VIHEAR
temple.
Although there are numerous research studies conducting both qualitative and
quantitative methods to investigate the impact of international tourism demand either in
ASEAN region or other developing countries, the study related Cambodia issue alone is not
yet exemplified sizably in the recent period. Accordingly, the study examines the impact of
economic and non-economic factors influencing to tourism demand for Cambodia during the
period of 1995 to 2015. Static and dynamic gravity approach based on fixed effect and
random effect as well as Arellano and Bond (1991) based GMM estimators will be applied.
0
200000
400000
600000
800000
1000000
1200000
1400000
To
uri
st A
rriv
als
(p
erso
n)
Asean East Asia Southern Asia
Oceania Europe Union Americas
The organization of the study is structured as follows: the 2nd section is to review
some empirical studies and journals is presented and methodology is presented in the 3rd
section. The 4th Section is to interpret the empirical results whereas the last section is to
conclude the whole finding.
2. Literature Reviews
On the application of the gravity model, on the one hand, were found to be numerous
in the context of international trade as well as migration issue. This approach was firstly
developed by Jan Tinbergen (1962) and adopted by Anderson and Van Wincoop (2003), Rose
(2000) and McCallum (1995) to study international trade flow. It is widely applied in the
context of foreign capital flow and migration flow, Bergstrand and Egger (2007),
Eichengreen and Tong (2007) and Head and Ries (2008) and Gil-Pareja et al. (2006) and
Karemera et al. (2000) respectively. Still, since international tourism is considered as service
flow, applying gravity model would be ideally interesting and considering such as a good
contribution. Yet, based on the research conducted by Alexander C. (2014) has mentioned
that the gravity model explains tourism flows better than goods trade for equivalent
specifications.
On the estimation of the factors influenced to tourism demand, on the other hand, is
met the maturity, there are a bunch of research but to propose static and dynamic gravity
models together is not yet found numerously. Most of them used international tourism
demand approach throughout time series and panel data model individually and separately.
Some employed a few economic variables to estimate its impact such as GDP per capital,
travel cost, exchange rate and CPI, Geoffrey I. Crouch (1994), Lim (1997), Li et al. (2005),
Song and Li (2008), Chukiat et al. (2010), Asrin Karimi (2015). It is worth notified that using
tourism demand model individually cannot capture the whole picture of tourist’s manner or
meaning that there would be existed other important factors that may influence to tourism
demand such as population, distance and relative cost of living or tourism zone (world
heritage or cultural tourism) or some others non-economic factors such as crisis, as well as
socio-economic political deadlock and country dummy variables as well.
Time series and panel data models are somehow applied. Ozan S. and Kadir K.
(2010), examined tourism demand in Turkey using panel gravity model based on fixed and
random effect estimator. Nuno C. L (2015), studied tourism demand in Portugal by employed
dynamic panel data based on the system GMM estimator. Accordingly, the author verified
that the dynamic model proves tourism demand is a dynamic process. Suparporn Sookmark
(2011) applied dynamic panel data based on Arellano and bond (1991) to estimate the factors
effecting international tourism in Thailand. The author stated that number of tourist arrivals in
the preceding year (t-1) is the main factor in determining their next visit (t+1). Asrin Karimi
(2015), studied tourism flow in ASEAN region, found that generalized Poisson regression
model is the best one to estimate long-run international tourism demand. H. Chantha (2015)
studied international tourism demand for Cambodia applying time series model, ARDL
approach. Worth notice, time series model individually cannot capture the information across
the country, there existed biased and lost some information based on its statistical and
econometrical inferences.
Moreover, all of these factors using in regression equation are found to be negatively
and positively correlated to tourism demand. Alexander C. (2014) found that the change in
exchange rate, bilateral trade as the share of business is positively affected to tourism. GDP
per capita and growth rate are also the crucial factors and showed the positive affected,
Chantha H. (2015) and Nuno Carlos Leitão (2015). Relative low prices of tourism in term of
cost of living and prices of goods and services have no effect, Roperto S. and Narae K.
(2014). Financial or global crisis has significantly effect to tourism, A Kusni et al. (2013).
Shortly, most of research studies have applied economic factors such GDP growth rate
and per capital, travel cost, tourism price to examine the impacts of tourism demand. Static or
dynamic estimator individually is used to investigate or somehow they applied both together.
Therefore, the contribution of this study is to extend gravity approach in which mostly used
in context of international trade to model international tourism demand and adopt more
crucial factors as well as applying country dummy variables as new contribution to the
previous related research in Cambodia.
3. Data and Methodology
Number of tourist arrivals from 32 origin countries during the period of 1995 to 2015,
used as an explained variable, was extracted annually from ministry of tourism, storage in
CEIC data manager, provided by Chiang Mai University (CMU). For the explanatory
variables such as GDP per capital and growth rate, exchange rate, inflation rate and total
population are imported from world development indicator (WDI), the World Bank.
Geography distance and transport cost are imported from www.Distancefromto.net and US
transport cost respectively. Relative cost of living index (RCL) is calculated based on Wong
et al., 2006 and Relative Production Index (RPI) methodology which is equated as follows:
lnRCLijt = ln [
CPIitCPIjt
⁄
EXRitEXRjt
⁄], and lnRPIijt = ln [
GRit
GRjt], where i is referred to a destination
country (Cambodia) and j is presented the origin country at time t
3.1 From Gravity Model to Cambodia’s International Tourism Demand Model
An adjustment of gravity model was firstly introduced by Tinbergen (1962) to analyze
the bilateral international trade derived from origin discovery of Newton’s theory of the law
of gravitation to analyze the attraction between two objects i and j. Therefore, this law is
equated as follows:
Fij = Gmβ1
imβ2j
di,j2 (1)
Where Fij is the force of gravitational between two objects i and j, m is the object i
and j, d is the distance between two objects i and j and G is universal gravitational constant.
From equation (1), we can rewrite it to the linear regression equation in the panel data
analysis with the natural logarithm and assuming G is equated to αi as follows:
ln(Fi,jt) = (G = αi) + μi + β1lnmit + β2lnmjt − γlnDij + εit (2)
Where Fi,jt is an explained variable, mit and mjt are vectors of explanatory variables
(normally measured as the economic size using a proxy of GDP between two countries), αi is
a constant term, μi is an unobserved country specific effect and εit is a normal distributed
error term assuming to be uncorrelated with μi. i and t is cross-section country and time
dimension respectively. From equation (2), it can be written into the international tourism
demand model as follows:
Yit = f(Xit, Zit) + αi + εit (3)
Where Yit is quantity of tourism demand from origin country i to destination country j
at time t, Xit is the set of explanatory variables and Zit is the set of control variables. From
equation (3), we can rewrite into international tourism demand for Cambodia based on the
concepts of gravity model with the nature of logarithm as follows:
lnTAit = αi + βlnGDPit + γlnGDPjt + δlnTCijt + θlnRCLijt + ϑlnPOit + γlnPOjt +
δlnDSijt + ηlnRPIijt + φk ∑ DVitni=1 + εit, where (4)
lnTAit is natural logarithm of international tourism arrivals from origin country j to
destination country i at time t
lnGDPjt is natural logarithm of GDP per capital of origin country j and 𝑙nGDPit is natural
logarithm of GDP per capital of destination country i at time t
lnTCijt is natural logarithm of travel cost from origin country j to destination country i at
time t
lnRCLijt is natural logarithm of relative cost of living index between origin country j and
destination country i at time t
lnPOjt is natural logarithm of total population of origin country j and lnPOit is natural
logarithm of total population of destination country i at time t
lnDSijt is natural logarithm of distance from origin country j to destination country i at
time t
lnRPIijt is natural logarithm of relative production index between origin country i and
destination country j at time t
∑ DVitni=1 are the set of dummy variables taking number 1 for the determined period and 0
otherwise. It is equated as follows:
φk ∑ DVitni=1 = φ1ASEAN1999 to 2015 + φ2Crisis2008/2009 + φ3AEC2015 +
φ4Election2003/ 2008/2013 + φ5eVisa2006 to 2015, (n = 1, 2, 3, 4, 5) (5)
Where ASEAN1999 to 2015 is referred to the period in which Cambodia joint ASEAN
region in 1997, Crisis2008/2009 is denoted the impact of global financial crisis in 2008 and
2009, Election2008 and 2013 is national election in 2008 and 2013, eVisa2006 to 2015 is denoted
the e-Visa starting to be launched. Yet, country specific dummy variables based different
regions are also included as follows:
∅k ∑ DCitki=1 = ∅1ASEAN Region + ∅2EU + ∅3East and South Asia + ∅4Oceanie +
∅5USA, (k = 1, 2, 3, 4, 5) (6)
Accordingly, from equation (4), (5) and (6), the study could be rewritten the new
regression equation of international tourism demand for Cambodia as follows:
lnTDit = αi + βlnGDPit + γlnGDPjt + δlnTCijt + θlnRCLijt + ϑlnPOit + γlnPOjt +
δlnDSijt + ηlnRPIijt + φ1ASEAN1999 to 2015 + φ2Crisis2008/2009 +
φ3AEC2015 + φ4Election2003/2008/2013 + φ5eVisa2006 to 2015 +
∅1ASEAN + ∅2EU + ∅3East and South Asia + ∅4Oceanie + ∅5USA + εit, (7)
Therefore, to investigate the factors which are potential in determining international
tourism demand for Cambodia, the equation (7) will be estimated throughout static and
dynamic models as follows:
3.2 Static and Dynamic Estimator
On the notification of Hsiao (2003, 2005), panel data sets are applied through three
different methods, namely pooled OLS, fixed effect (FE) and random effect (RE) estimators.
Therefore, the regression equation of static panel data is equated as follows:
Yit = αi + β1Xit + β2Wit + vi + εit, i = 1, … , N and t = 1, … , T (8)
Where μit = vi + εit is the country specific effect
Hsiao (2003), in pooled OLS estimator, takes into account the country specific effect;
accordingly panel data models based on FE and RE estimator use to eliminate those problems
by considering the ideas as follows.
FE estimator is assumed that the slops are common and differ in intercept and allowed
for unobservable country heterogeneity whereas in RE estimator, considered unobservable
country heterogeneity effect but variation across the entities (unobserved individual effect,
(αi) are random and uncorrelated with independent variables, followed by normal
distribution. Unlike FE, RE is incorporated these effects into error term which assumed to be
uncorrelated with dependent variable (Hsiao, 2003). It is worth noting that, since time-
invarying variables such as distance and common languages as well as religion were removed
in FE estimator for which leaded to be less efficiency. Accordingly, to eliminate that issue,
RE estimator takes into account.
Therefore, to select whether FE or RE estimator is appreciated, Hausman (1978) test
was adopted to detect under the null hypothesis of RE is better where conversely FE is better
for the alternative one. Hausman (1978) test is equated as follows:
H = (β̃1,RE − β̂1,FE)[cov̂(β̃1,RE − β̂1,FE)]−1
(β̃1,RE − β̂1,w), where β1 corresponds to time-
varying regressor.
Static panel data model is produced bias, inconsistence and misleading inference
when the existence of endogeneity in the independent variables based on the finding of
Baltagi, Egger and Pfaffermayr (2003). Yet, to control such issue as well as the lagged
dependent variable using as instrument value, dynamic estimator or dynamic FE estimator
taking the lagged dependent variable of generalized method of moment (GMM) was
proposed. Therefore, the dynamic panel data model is equated as follows:
Yit = ∑ αiYi,t−p +pj=1 β1Xit + β2Wit + vi + εit, i = 1, … , N and t = 1, … , T (9)
If indeed there presented the correlation between lagged dependent variable and
country specific fixed effect (μit), then it will lead to be biased estimators in panel data model
(Nickell, 1981). This bias will be disappeared only if time periods go to infinity (T → ∞),
(Nickell, 1981). To remove this, Arellano-Bond (1991) had suggested a GMM estimator
taking into account the dynamic lagged of depended variable to be uncorrelated with error
term (absence of autocorrelation).
More importantly, GMM is the efficiency and consistency technique in removing the
problem appeared in FE and RE estimator by using instrument variables to avoid the
endogenous issue whereas the moment condition is the orthogonality conditions within
lagged dependent variable and error terms in panel data model. The consistency of GMM
estimator gives the fact that E(Δεit, Δεit−2) = 0 and it works poorly in short panel (meaning
that, N and T is small, but works efficiency in the case of N is big and T is small), Blundell
and Bond (1991). Furthermore, on the usage of moment conditions suggested by Arellano
and Bond (1991) rising its number when the time periods T is increased. Thus, the estimation
needs accordingly to test the over-identification restrictions by Sargan test.
4. Empirical Results and Discussion
The empirical results in the Table 1 are estimated throughout static panel data models
by taking into account international tourist arrivals (TA) as a dependent variable. The
diagnostic statistical tests such as Sargan, Hausman test, AR process were detected. The
results indicated that international tourist arrivals to Cambodia are mostly attracted by an
increasing of income per capital, population growth and the others non-economic factors such
as crisis and national election within the destination country. FE and RE indicated that GDP
per capita of both destination and origin countries are significantly and positively influenced
to TA where relative production index (RPI) and relative cost of living (RCL) are negatively
affected and significant at 5%.
Table 1: Empirical results from static gravity models with and without robust standard
error
Variables
Static Estimator without Robust
Standard Error
Static Estimator with Robust Standard
Error
FE RE FE RE
Constant 68.2093**
(2.06)
84.8416***
(2.76)
68.2093***
(3.67)
84.8416***
(2.76)
Explanatory Variables
lnGDPit 3.9969***
(5.21)
4.0212***
(5.24)
3.9969***
(7.12)
4.0212***
(6.59)
lnGDPjt 0.4322***
(2.35)
0.7109***
(5.55)
0.4322
(1.39)
0.7109***
(4.61)
lnPOit -7.1514***
(-3.32)
-7.2054***
(-3.39)
-7.1514***
(-5.67)
-7.2054***
(-6.41)
lnPOjt 1.5176
(2.18)**
0.8302***
(7.83)
1.5176
(1.23)
0.8302***
(7.00)
lnRPIijt -0.04158
(-1.36)
-0.0364
(-1.20)
-0.0416***
(-2.20)
-0.0364
(-1.92)
lnRCLijt -0.0487*
(-1.62)
-0.0470**
(-1.70)
-0.0487
(-1.16)
-0.0470**
(-1.28)
lnTCijt 0.4789***
(3.82)
0.4686***
(3.71)
0.4789***
(3.12)
0.4686***
(2.91)
lnDSijt 0.2575
(0.23)
-0.6166
(-1.43)
0.2575
(0.83)
-0.6166
(-1.61)
ASEAN1997 to 2015 0.1026
(0.94)
0.0088
(0.81)
0.1026
(0.98)
0.0088
(0.87)
Crisis2008/2009 -0.1715
(-1.38)
-0.173
(-1.39)
-0.1715***
(-2.07)
-0.173
(-2.12)
Election2008 and 2013 0.1559**
(1.93)
0.1559**
(1.91)
0.1559***
(2.33)
0.1559**
(2.35)
eVisa2011−2015 -1.5709***
(-12.35)
-1.5702***
(-12.24)
-1.5709***
(-11.89)
-1.5702***
(-11.80)
Country Dummy Variables
ASEAN Region ---- 0.8134
(0.69) ----
0.8134
(0.72)
East and South Asia ---- 0.3305
(0.37) ----
0.3305
(0.44)
EU 0.2350
(0.42)
0.2866
(0.51) ----
0.2866
(1.45)
Oceania ---- 1.6295**
(1.91) ----
1.6295**
(4.39)
USA 0.7541***
(0.87) ----
0.7541***
(2.58)
R square (R2) 0.3671 0.3671 0.3671 0.6335
F-statistic 46.85***
[0.0000]
46.85***
[0.0000] -----
728539.77***
[0.0000]
Breush-Pagan Testa ----- 2286.48***
[0.0000] -----
2286.48***
[0.0000]
Hausman Test 23.31**
[0.0381]
23.31**
[0.0381] ----- -----
Sigma_e 1.8933 0.8109 1.8933 0.81086
Sigma_u 0.5354 0.5354 0.5354 0.5354
Rho 0.92595 0.6964 0.92595 0.6964
Note: The notification sign of *, ** and *** denote the significant level of 10%, 5% and 1% respectively. The
value inside the parenthesis () and [] is referred to z-statistic for RE and t-statistic for FE and p-value
respectively. RE and FE models are estimated using with and without robust standard errors.
Source: Computer calculation
Distance (DS) between both countries is positively correlated but not significant. With
the respect to dummy variables using as binary option are found to be either negative or
positive and significant or insignificant. Crisis and eVisa is negatively affected to TA whereas
as a member of ASEAN and national election is positively associated but not significant.
More importantly, taking into consideration the country specific dummy variables of
five different regions, namely ASEAN, East and South Asia, EU, USA and Oceania indicated
that these variables explained well in RE estimator and conversely due to multicollinearity,
FE could not estimate its coefficients. The diagnostic tests indicated that Breush-Pagan LM
test for RE can be rejected the null hypothesis at 5% level of significant whereas Hausman
test for FE is appreciated and rejected the null hypothesis at 5% level with the statistical value
of 23.31. With the respect to R square (R2) statistic, RE with robust standard error is explained
better rather than other three models did.
The empirical results from dynamic panel data based on GMM estimator both one and
two step system, employed GDP, GDP per capital, crisis and election as instrument variables
are demonstrated in the Table 2. For five different regions, TA is mostly attracted by income
of the origin country, travel cost, production index and the others non-economic factors as
demonstrated as similar as the static models. The uncertainty of national election during the
period of 2008 and 2013 is crucially and positively affected. The diagnostic tests indicated
that Wald chi-squares of all models can be rejected the null hypothesis at 1% level of
significant, meaning that the sample observations were fitted perfectly to the models. The
country specific dummy variables based on five different regions are removed due to
multicolinearity problem. Sargan test for over restriction of GMM without robust standard
error is 399.4379 and 29.2557 and rejected the null hypothesis at 1% level for one and two
step system respectively. Arellano and Bond test for autocorrelation with the AR(1) is -2.816
and -2.7853 with 1% level of significant for two step GMM without robust standard error and
one step GMM with robust standard error respectively and AR(2) process is -2.4456 and -
2.6285 with 1% level of significant for two step GMM without robust standard error and one
step GMM with robust standard error respectively as well. Based on these four different
models demonstrated that the lagged depended variable is positively associated and
significant at 1% level and GDP per capital of the origin and destination country, population
growth rate (PO) of the origin country and travel cost (TC) are positively impacted and
significant at 1% level whereas oppositely population growth rate of the destination country
and relative production index (RPI) are negatively impacted and significant at 1%. Most of
dummy variables are better explained the movement of tourist arrivals to Cambodian market.
Table 2: Empirical results from dynamic gravity models with and without robust standard
error
Variables
Dynamic Estimator based GMM without
Robust Standard Error
Dynamic Estimator based GMM with
Robust Standard Error
One Step System Two Step System One Step
System Two Step System
Constant term (c) 105.2886***
(3.39)
65.1977**
(1.79)
106.6571***
(3.48)
65.1977
(0.16)
Explanatory Variables
lnTAit−1 0.2794***
(7.36)
0.2746***
(25.94)
0.3058***
(6.28)
0.2746
(0.20)
lnTAit−2 ----- ----- -0.0493
(-1.10) -----
lnGDPit 5.1537***
(8.66)
4.8697***
(18.95)
5.0825***
(7.61)
4.8697
(0.72)
lnGDPjt 0.3901
(1.57)
0.0812
(0.20)
0.3039
(0.89)
0.081
(0.01)
lnPOit -10.6085***
(-5.46)
-10.8667***
(-10.61)
-9.9417***
(-4.79)
-10.8667
(-0.56)
lnPOjt 2.2453**
(1.71)
5.0747**
(1.76)
1.6261
(1.42)
5.0747
(0.33)
lnRPIijt -0.0789***
(-4.04)
-0.0782***
(-9.26)
-0.0818***
(-4.23)
-0.0782
(-0.30)
lnRCLijt -0.0302
(-1.02)
-0.0739**
(-1.21)
-0.0118
(-0.41)
-0.0739
(-0.03)
lnTCijt 0.0399
(0.51)
0.0507**
(1.66)
0.0452
(0.79)
0.0507
(0.08)
Dummy Variables
ASEAN1997 to 2015 0.1553**
(1.93)
0.1658***
(3.61)
0.1013
(1.59)
0.1658
(0.06)
Crisis2008/2009 -0.2206***
(-3.02)
-0.2101***
(-8.18)
-0.2614***
(-9.30)
-0.2101
(-0.21)
Election2008 and 2013 0.2526***
(5.67)
0.2246***
(8.92)
0.2559***
(3.76)
0.2246
(0.53)
eVisa2011−2015 -1.4458***
(-18.89)
-1.3928***
(-22.89)
-1.3697***
(-9.30)
-1.3928
(-2.26)
Number of Instrument 202 202 201 202
Wald Chi2 1171.83***
[0.0000]
9379.31
[0.0000]
1065.55***
[0.0000]
198.34***
[0.0000]
Sargan Test 399.4379***
[0.0000]
29.2557***
[0.0000] ----- -----
Arellano-
bond Test
AR(1) -0.3995
[0.6895]
-2.816***
[0.0049]
-2.7853***
[0.0053]
-0.3995
[0.6895]
AR(2) -0.3269
[0.7437]
-2.4456***
[0.0145]
-2.0285***
[0.0425]
-.03269
[0.7437]
Note: The notification sign of *, ** and *** denote the significant level of 10%, 5% and 1% respectively. The
value inside the parenthesis is referred to z-statistic and in the square parenthesis () and [] is indicated the t-
statistic and p-value respectively. GMM estimators: Instrument variables GDPi, GDPj, RPIi, RCLj, FC and
Election, Crisis. Sargan is a test of over-identifying restrictions in GMM estimation. Arellano–Bond test for
analyzing the autocorrelation existence of second order (p-value) based AR(2).
Source: Computer calculation
Shortly, in accordance to the above empirical results obtained from static and dynamic
models indicated that dynamic estimator seems to be well performance rather than those of
static did. Thus, it produced the better results in dynamic process for the sample observations.
5. Conclusion and Policy Implication
International tourist arrivals to Cambodian market has determined by the vital
economic and non-economic factors. Panel data with 32 cross-sections and covered 22 years
(1995 – 2015) are used to estimate its impact and the dynamic correlation throughout static
and dynamic estimators based on the idea of gravity approach which was developed from
international trade flow. Primarily, static gravity model extending to international tourism
demand one is estimated using FE and RE estimator. Subsequently, the dynamic estimators
are proposed by adding the lagged dependent variable as the dynamic regressor. The results
from both static and dynamic estimators indicated that international tourist arrivals into
Cambodia are empirically determined by population, distance, exchange rate, economic
growth and per capital as well as the tourism flow of the previous year. It is likely indicated
that tourism who come to visit in Cambodia in the current period is considered as the catalyst
to attract others tourist to come and visit. Simply, if one country have been going to other
countries, the others is willing as well. Population size, income per capita of both countries,
tourism price and cost of living and the development of marketing strategy of the destination
country is also the main factors to be impacted.
By understanding those influencing factors, Cambodia’s government should firstly
maintain the stability of price and cost of living since they are very crucial for encourage
tourism attention as recently Cambodia approved from lower income country to middle lower
income one. Secondly, adopting policies-related marketing during the main national events
such as Khmer New Year or Water Festival should be considered. Therefore, the economic
policy implication toward tourism sector is principally robust with the regards to market
diversification via-a-via income per capital of origin tourist, cost of living and the rising of
population but the effect is not uniform across countries.
References
Asrin, K., Pouya, F., & Khalid, A. R. (2015). Modeling and forecasting of international
tourism demand in ASEAN countries. American Journal of Applied Sciences, 12(7),
479-486.
Culiuc, A. (2014). Determinants of international tourism.
Kusni, A., Kadir, N., & Nayan, S. (2013). International tourism demand in Malaysia by
tourists from OECD countries: A panel data econometric analysis. Procedia
Economics and Finance, 7, 28-34.
Anderson, J. E., & Van Wincoop, E. (2003). Gravity with gravitas: a solution to the border
puzzle. the american economic review, 93(1), 170-192.
Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations. The review of economic
studies, 58(2), 277-297.
Blundell, R., & Bond, S. (2000). GMM estimation with persistent panel data: an application
to production functions. Econometric reviews, 19(3), 321-340.
Baltagi, B. H., Bresson, G., & Pirotte, A. (2003). Fixed effects, random effects or Hausman–
Taylor?: A pretest estimator. Economics letters, 79(3), 361-369.
Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel
data models. Journal of econometrics, 87(1), 115-143.
Hausman, J. A., & Taylor, W. E. (1981). Panel data and unobservable individual
effects. Econometrica: Journal of the Econometric Society, 1377-1398.
Hor, C. (2015). Modeling International Tourism Demand in Cambodia: ARDL Model.
Chaiboonsri, C., Sriboonjit, J., Sriwichailamphan, T., Chaitip, P., & Sriboonchitta, S. (2010).
A Panel Cointegration Analysis: An Application to International Tourism Demand of
Thailand. Annals of the University of Petrosani Economics, 10(3).
Chaitip, P., Chaiboonsri, C., & Rangaswamy, N. (2008). A Panel Unit Root and Panel
Cointegration Test of the Modeling International Tourism Demand in India. VOL. VIII
PART I, 95.
Chaiboonsri, C., Chaitip, P., Rangaswamy, N. (2006). Modelling International Tourism Demand in
Thailand. Annals of the University of Petroşani, Economics, 9(3), 2009, 125-146
Muchapondwa, E., & Pimhidzai, O. (2011). Modelling international tourism demand for
Zimbabwe. International journal of business and social science, 2(2).
Hsiao, C. (2014). Analysis of panel data (No. 54). Cambridge university press.
Crouch, G. I. (1994). The study of international tourism demand: a review of
findings. Journal of Travel research, 33(1), 12-23.
Martínez-Zarzoso, I., Felicitas, N. L. D., & Horsewood, N. (2009). Are regional trading
agreements beneficial? Static and dynamic panel gravity models. The North American
Journal of Economics and Finance, 20(1), 46-65.
Hausman, J. A. (1978). Specification tests in econometrics. Econometrica: Journal of the
Econometric Society, 1251-1271.
Hausman, J. A., & Taylor, W. E. (1981). Panel data and unobservable individual
effects. Econometrica: Journal of the Econometric Society, 1377-1398.
Karemera, D., Oguledo, V. I., & Davis, B. (2000). A gravity model analysis of international
migration to North America. Applied Economics, 32(13), 1745-1755.
Keum, K. (2010). Tourism flows and trade theory: a panel data analysis with the gravity
model. The Annals of Regional Science, 44(3), 541-557.
Kareem, O. I. (2008). A panel data analysis of demand for tourism in Africa. Ibadan Journal
of Social Sciences, Forthcoming.
Saray, M., & Karagöz, K. (2012). Determinants of tourist inflows in Turkey: Evidence from
panel gravity model. Uluslararası Yönetim İktisat ve İşletme Dergisi, 6(11), 33-46.
Leit, N. C. (2015). Portuguese tourism demand: a dynamic panel data analysis. International
Journal of Economics and Financial Issues, 5(3).
Shareef, R., & McAleer, M. (2007). Modelling the uncertainty in monthly international
tourist arrivals to the Maldives. Tourism Management, 28(1), 23-45.
Sookmark, S. (2011). An analysis of international tourism demand in Thailand.
Nonthapot, S., & Ueasin, N. International Tourism Demand in the Greater Mekong Sub-
Region: A Panel Granger Causality Approach.
Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica: Journal of
the Econometric Society, 1417-1426.
Wong, K. K., Song, H., & Chon, K. S. (2006). Bayesian models for tourism demand
forecasting. Tourism Management, 27(5), 773-780.
Su, Y. W., & Lin, H. L. (2014). Analysis of international tourist arrivals worldwide: The role of world
heritage sites. Tourism Management, 40, 46-58.
Eilat, Y., & Einav, L. (2004). Determinants of international tourism: a three-dimensional
panel data analysis. Applied Economics, 36(12), 1315-1327.
Appendices
Appendix I: Selected origin countries arrival to Cambodia using in the estimated
regression
Regions Countries
ASEAN Brunei, Indonesia, Laos, Malaysia, Myanmar,
Philippines, Singapore, Thailand, Vietnam
East and Southern Asia China, Hong Kong, Taiwan, Japan, Korea, India,
Sri Lanka
European Union
Denmark, Finland, Norway, UK, Austria,
Belgium, France, Germany, Netherlands,
Switzerland, Italy, Spain
American USA and Canada
Oceania Australia and New Zealand
Appendix II: The hypothesized signs of all the variables using the static and dynamic panel
data model
Variable Description Expected sign
Dependent and Independent variables
TAijt Number of international tourist arrivals from origin countries to
Cambodia ----
TAijt−1 Lagged dependent variable using as endogenous regressor in
dynamic regression +
GDPit Real gross domestic product (GDP) per capital of destination
country i +
GDPjt Real gross domestic product (GDP) per capital of origin country j +
TCijt Travel cost from origin country j to destination country i +/−
RCLijt Relative cost of living index between origin country j and
destination country i +/−
POjt Total population of origin country j +
POjt Total population of destination country i +
RPIijt Relative production index between origin country i and
destination country j +/−
DSijt Distance in kilometers between the capital cities of origin and
destination +/−
Dummy variables
ASEAN membership
Selected since Cambodia became one of the member of ASEAN
region and takes the value 1 in the determined period and 0
otherwise
+
GCrisis It is representing the influencing of the global economic crisis
takes 1 during the crisis period 2008–2009 and 0 otherwise −
e-Visa Taking the year since Cambodia started to launch e-visa in 2011
and takes the value 1 in the determined period and 0 otherwise +
Election
Dummy Variable started between three different mandates of
national election and takes the value 1 in the determined period
and 0 otherwise
+/−
Note: Signs (+) and (−) correspond to the expected positive and negative effects on the impact on the magnitude
of international tourism flow capturing from both theoretical framework and empirical research publications.
Appendix II: The hypothesized signs of all the variables using the static and dynamic panel
data model (Cont.)
Variable Description Expected sign
Country Specific Region Dummy Variables
ASEAN Number of country in ASEAN takes the value 1 in the
determined country of those who is in the region and 0 otherwise
+/−
East and South Asia
Number of country in East and South Asia takes the value 1 in
the determined country of those who is in the region and 0
otherwise
+/−
EU Number of country in EU takes the value 1 in the determined
country of those who is in the region and 0 otherwise
+/−
Oceania Number of country in Oceania takes the value 1 in the
determined country of those who is in the region and 0 otherwise
+/−
USA Number of country in USA takes the value 1 in the determined
country of those who is in the region and 0 otherwise
+/−
Note: Signs (+) and (−) correspond to the expected positive and negative effects on the impact on the magnitude
of international tourism flow capturing from both theoretical framework and empirical research publications.
Appendix II: Pearson's correlation coefficient of all variables
Series LnTA LnGDPi LnGDPj LnPOi LnPOj LnRPI(ij) LnRCL(ij) LnFC LnDS
LnTA 1.0000
LnGDPi 0.2644 1.0000
LnGDPj 0.1048 0.0559 1.0000
LnPOi 0.2621 0.9871 0.0574 1.0000
LnPOj 0.5006 0.0354 -0.4411 0.0296 1.0000
LnRPI(ij) 0.0128 0.1058 0.5234 0.1225 -0.2377 1.0000
LnRCL(ij) -0.1029 -0.0087 -0.6703 -0.0142 0.2136 -0.3777 1.0000
LnFC 0.2496 0.9405 0.0489 0.9006 0.0300 0.0877 -0.0085 1.0000
LnDS -0.0004 -0.0427 0.7357 0.7357 -0.0392 -0.0705 0.4448 -0.0486 1.0000
Note: Pearson’s correlation (r) indicated perfect and imperfect correlation between one variable to others one whereas -1 < r < +1. From above table indicated that Poi is
correlated with GDPi and it is correlated with FC. DS is correlated with GDPi and Poi while POi did with FC. Thus, in estimated regression, correlation’s variables will be
dropped to avoid the problem of multicollinearity which leads to be inconsistency and bias results and leads to higher t-statistic.